Evidence of Industry Scale Effects on Audit Hours, Billing Rates, and Pricing
SIMON DEKEYSER, KU Leuven ANN GAEREMYNCK, KU Leuven
MARLEEN WILLEKENS, KU Leuven and BI Norwegian Business School†
* Accepted by Jeffrey Pittman. The authors acknowledge the useful comments of seminar participants at KU Leuven, Skema Business School, Lille, and Vrije Universiteit Amsterdam. We would like to thank the editors and two anonymous reviewers for their excellent suggestions. We are indebted for constructive feedback on prior versions of the article to Liesbeth Bruynseels, John Christian Langli, and W. Robert Knechel, as well as to participants at the EARNET conference, the EAA Annual Congress, the International Symposium on Audit Research (ISAR), the 4th workshop on Audit Quality of the EIASM, and the AAA Auditing Section Midyear Conference in New Orleans, Louisiana. Simon Dekeyser gratefully acknowledges financial support from the Research Foundation – Flanders (FWO).
† Corresponding author: marleen.willekens@econ.kuleuven.be
Accepted Article
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite
10.1111/1911-3846.12460
Evidence of Industry Scale Effects on Audit Hours, Billing Rates, and Pricing
Abstract
Using a proprietary dataset consisting of all private firm audit engagements in 2000 from one Big 4 firm in Belgium, we investigate: (1) whether audit office industry scale is associated with a reduction of total, partner and staff audit hours and thus with efficiency gains triggered by organizational learning from servicing more clients in an industry;
and (2) whether the extent of efficiency pass-on from the auditor to its clients depends on the audit firm’s market power. We find that auditor office industry scale is associated with efficiency gains and a reduction of the variable costs (i.e., fewer total audit hours, partner hours, and staff hours), ceteris paribus. Our results also suggest that, on average, realized efficiencies are entirely passed on as evidenced by a non-significant effect of auditor industry scale on the auditor’s billing rate. Furthermore, we find that the extent of the efficiency pass-on decreases with the market power of the audit firm in the industry market segment as we document a higher billing rate for auditors with high market power (versus low market power). In addition, we find that the lower audit hours associated with auditor industry scale do not compromise audit quality.
Keywords: industry knowledge, audit efficiency, industry market power, efficiency pass-on
JEL-classification: M41, M42
Accepted Article
Evidence of Industry Scale Effects on Audit Hours, Billing Rates, and Pricing
1. Introduction
In this article, we study the effects of audit office industry scale derived from specializing and investing in a particular client industry on audit hours, audit fees, and billing rates (fee per hour).1 We argue that a larger audit office industry scale results in more industry-specific knowledge, which is an important attribute for attaining a comprehensive understanding of the client’s operating environment and accounting practices. Theoretically, higher industry knowledge could result in two distinct outcomes: a higher-quality audit service (Reichelt and Wang 2010) or an audit service that has the same level of quality but is more efficient (Casterella et al. 2004; Bills, Jeter and Stein 2015). The learning curve paradigm from management science literature suggests that the cumulative expertise gained from servicing more clients in a specific industry (i.e., a higher industry scale) results in enhanced
organizational knowledge about the client’s industry, and hence in more efficient audits (i.e., fewer audit hours).2 We therefore hypothesize that auditor industry scale is negatively associated with the number of audit hours performed on engagements in that industry (Hypothesis 1).
Next, we investigate whether efficiencies realized from industry scale are passed on from the auditor to its clients and whether the extent of efficiency pass-on depends on the audit firm’s market power in the relevant audit market (industry) segment. Theoretical models of the relationship between pricing and organizational learning (Cabral and Riordan 1994; Besanko et al. 2010) predict a full pass-on of efficiencies and even pricing below marginal cost. The latter phenomenon is known as predatory pricing or price undercutting (Cabral and Riordan 1994). Obtaining a client allows the firm to go further down its learning curve, achieving additional efficiency gains.
At the same time, rivals are prevented from achieving similar learning effects, which increases the firm’s
1In this paper we refer to the actual fee per hour (staff or partner) for effort performed on the engagement as the “billing rate.”
Therefore, the billing rate reflects how much the audit firm is compensated for each hour invested in the engagement.
2The learning paradigm has also been applied to manufacturing (Argote and Epple1990) as well as to service firms (Booneet al.
2008; Bohet al. 2007; Reaganset al. 2005; Pisanoet al. 2001; Darret al. 1995). Cahanet al. (2011) argue that industry
specialization can be achieved either through increased quality or decreased cost and that audit firms are more likely to pur sue the latter strategy in small client settings.
Accepted Article
competitive advantage. Consequently, in the short run, a firm is willing to pass on efficiencies to its clients because having more clients results in a stronger competitive position. In the long run, however, such initial predatory pricing may drive competitors out of the market resulting in non-competitive markets with fewer suppliers, higher market power for the remaining competitors, and higher prices (Dasgupta and Stiglitz 1988; Besanko et al. 2010).
Besanko et al. (2010) show that the extent of pass-on decreases as firms become more asymmetric, which occurs for example when firms have differential market power.3 We therefore predict that the extent of efficiency pass-on is conditional on an auditor’s market power: the greater an auditor’s market power, the smaller the extent of efficiency pass-on (Hypothesis 2).
Prior research on auditor industry knowledge has devoted considerable attention to the audit pricing and quality effects of market share-based industry specialization, i.e., an auditor’s industry knowledge relative to its competitors (Craswell et al. 1995; Ferguson et al. 2003; Francis et al. 2005; Gul et al. 2009; Reichelt and Wang 2010; Cahan et al. 2011). The results are mixed, but a majority of studies support the claim that market share-based industry specialists charge higher audit fees, provide higher quality, and perform more audit hours (Bae et al. 2016).
On the contrary, research on the effects of auditor industry scale (or an auditor’s internal, i.e., within audit firm, industry focus) is relatively scarce. Fung et al. (2012) study audit office industry scale and report a negative association with audit fees which they attribute to efficiency gains. However, Fung et al. (2012) do not use audit hours data in their analysis (but audit fees), and therefore are presenting indirect evidence of efficiency gains. In this paper, we add to the literature by examining the effect of industry knowledge arising from a larger office industry scale on audit hours thereby providing a direct test to differentiate between these two theories.
As indicated, we also study whether realized efficiency gains are passed on to the audit clients. Prior literature in this area is scarce and again the inferences are mainly based on audit fee models (Casterella et al. 2004;
Fung et al. 2012; Bills et al. 2015). As these fee studies do not account for audit hours, they are unable to unravel the source of the documented fee discounts, which could be attributable to: (i) more or less efficiency pass-on; or (ii) differences in the size of realized efficiency gains. A notable exception is a study by Bae et al. (2016) which does use an audit hours and billing rate model to test industry knowledge effects, but the focus is on market-share based
3 Competing audit firms are deemed ‘asymmetric’ (or different) when they are dissimilar to one another, for instance, in terms of know-how or market share.
Accepted Article
measures of industry specialization instead of within-audit firm industry scale effects. Both proxies for industry specialization capture different aspects of industry expertise, and market share based proxies of industry specialization are likely to also capture to some extent an auditor’s market power. We add to the literature by examining whether efficiency gains (i.e., reduction in audit hours) from audit office industry scale, if realized, are indeed passed on to audit clients via lower billing rates and whether the extent of the efficiency pass-on depends on the audit firm’s market power.
To test our hypotheses, we use a unique proprietary dataset of all (895) private firm audit engagements from one Belgian Big 4 firm for the year 2000, containing audit hours as well as audit fees. We augment this data with the financial information contained in the clients’ annual reports. In line with prior research, we measure and test scale at the audit office level as knowledge transfer from the office to the national level is imperfect because industry-specific knowledge is tacit and difficult to codify (Francis et al. 2005). We specify two measures of office industry scale: (1) a continuous measure based on Fung et al. (2012) capturing the number of clients in each office industry; and (2) a dummy variable equal to one for the three industries in which the audit office has invested most of its resources. Finally, we measure an auditor’s market power within the relevant industry market segment using the absolute market share distance to the closest competitor (Numan and Willekens 2012; Bills and Stephens 2016).
Our results show that audit office industry scale is negatively associated with the number of audit hours spent on engagements. These results are in line with the learning curve paradigm: servicing more clients in a particular industry, or having a higher scale, allows a firm to go down on the learning curve thereby increasing audit efficiency and decreasing the cost of production. The results also illustrate that, on average, realized efficiencies are entirely passed on to clients via fee discounts as evidenced by a significantly lower audit fee but an insignificant difference in the billing rate. The results further show that when the audit firm’s market power is strong, the extent of the efficiency pass-on decreases. When market power is low, efficiency gains due to scale are entirely passed on to clients, but the extent of efficiency pass-on decreases when market power is high resulting in a higher billing rate.
Supplementary analysis reveals that our measures of audit office industry scale are not associated with accruals–
based earnings management and thus do not result in a decrease of the audit quality supplied (Dutillieux 2009;
Reichelt and Wang 2010; Cahan et al. 2011; Willekens et al. 2017).
We contribute to the literature in the following ways. First, we contribute to the scarce literature on audit office industry scale by examining effects on audit hours instead of audit fees. To our knowledge, no prior study
Accepted Article
directly examines auditor industry scale effects on audit hours. Second, we also investigate whether the scale effects vary for different types of audit hours and find efficiency gains at the staff level (documented for both industry scale measures adopted) and at the partner level (documented for one industry scale measure). As such we contribute to the growing literature of research on individual partner effects in auditing research (Taylor 2011; Knechel et al.
2015). Third, by examining auditor industry scale effects on billing rates and audit fees, we evaluate whether the efficiencies are passed on to clients. Note that this kind of analysis could not be performed in Bills et al. (2015) and Fung et al. (2012) due to lack of audit hours data. Note further that prior research has investigated the effect of demand-side characteristics such as (relative) client size on audit fees (Casterella et al. 2004; Fung et al. 2012; Bills and Stephens 2016) or focus on the homogeneity of the industry (Bills et al. 2015) while we take a supply-side perspective by focusing on the firm’s market power and the billing rate. Fourth, different from Fung et al. (2012) and Bae et al. (2016), we study audit clients in the private firm segment of the audit market. While some prior studies analyze audit production function in public firms, little is known about audit production in the private client segment. Finally, we add to the debate on the measurement of auditor industry specialization (Minutti-Meza 2013;
Audousset-Coulier et al. 2015). Our results show a different audit hour and audit fee effect of our portfolio-based measures of industry expertise (such as our audit office industry scale measures) versus our market-share based measure of industry expertise (the market leadership measure often used in the literature and included as a control variable in our models). This corroborates our argument that market share-based and portfolio share-based measures of industry specialization most likely capture two distinct aspects of auditor industry knowledge.
The remainder of this article is organized as follows: Section 2 provides an overview of the literature and develops the hypotheses to be tested. Section 3 describes the research design and section 4 discusses the sample and the data. In section 5, we report the main results. Section 6 reports some sensitivity checks. We discuss the
conclusions and the limitations of this study in section 7.
2. Previous literature and hypotheses
Prior literature on auditor industry knowledge: Economies of scale or quality premium?
It is well established in the auditing literature that industry-specific (accounting) knowledge is a valuable attribute for the execution of audit engagements (Danos et al. 1989; Craswell et al. 1995; Balsam et al. 2003; Francis et al.
2005; Malhotra and Morris 2009). The auditing literature shows that auditor industry knowledge is associated with
Accepted Article
higher fees (Ferguson et al. 2003; Francis et al. 2005; Craswell et al. 1995; Cahan et al. 2011; Numan and Willekens 2012; Bills et al. 2015), higher audit quality (Balsam et al. 2003; Gul et al. 2009; Reichelt and Wang 2010),4 and even higher audit effort (Bae et al. 2016). Other studies report decreased audit fee premiums or even fee discounts associated with industry knowledge, depending on several factors: the type of audit firm (DeFond et al. 2000), the proportion of clients audited in that industry (Cahan et al. 2011), the client bargaining power (Fung et al. 2012;
Casterella et al. 2004), and industry characteristics (Bills et al. 2015). These results are mainly attributed to efficiency gains, although none of these studies actually empirically examine audit hours.
A relevant observation in this light is that prior studies typically use market share-based proxies of industry specialization to capture enhanced auditor industry expertise, such as industry leadership. Market share-based measures of industry specialization, however, represent how well an audit firm “has differentiated itself from its competitors in terms of market share within a particular industry” (Neal and Riley 2004, 170; italics in original), and thus, by construction, also pick up an auditor’s location (and thus market power) relative to other competitors in the market (Numan and Willekens 2012). Market share-based measures of industry specialization are not necessarily suitable measures of auditor industry scale derived from specializing and investing in a particular client industry. At the extreme, using market share-based measures of industry specialization, an auditor can be deemed an industry expert with only one or two large clients in its portfolio if there are only a few clients from that industry present in the market segment (Neal and Riley 2004). Consistent with this observation, Fung et al. (2012) introduce a different measure to capture auditor industry scale. Note that Fung et al. (2012) do not use audit hours and therefore present indirect evidence of efficiency gains by examining audit fee discounts. Audit fee discounts can occur for two theoretically distinct reasons. First, a larger scale allows a firm to divide the fixed costs, such as training and office rent, over more clients, thereby decreasing the average cost and subsequently the audit fee. Second, a larger scale could also increase audit efficiency, thereby decreasing the variable production cost of the audit service and hence the audit hours necessary to execute the engagement. Testing the effects of scale using an audit fee model as done in
4Studies that investigate industry knowledge at the partner level—rather than the firm or office level—find that higher industry knowledge leads to a better understanding of possible material misstatements in pre-audited financial statements, resulting in a higher likelihood of those misstatements being detected and resolved (Hammersley 2006), and a better understanding of a client’s business environment (Eichenseher and Danos 1981), as well as of industry-specific accounting practices (Taylor 2000).
Accepted Article
Fung et al. (2012) precludes disentangling these two competing explanations, as it is unclear whether the observed negative audit fee effects stem from lower variable costs (i.e., more efficient audits) or the distribution of fixed costs over more engagements (i.e., lower billing rate). Our study extends the literature by simultaneously examining audit hours and billing rates.
Economies of industry scale and audit efficiency (H1)
The first objective of this study is to test the effect of office industry scale on audit hours. We rely on two non- mutually exclusive theories to predict the direction of this effect: economies of scale and organizational learning.
Economies of scale refer to the decreasing cost per unit of a product or service as the scale of operation increases (Besanko et al. 2010). Economies of scale arise primarily from the division of fixed costs over more units leading to lower average costs and occur often in capital-intensive industries. Hence, economies of scale primarily predict a negative effect of office industry scale on the billing rate as each engagement has to cover a smaller component of the fixed costs. However, this theory does not necessarily impact the variable costs, such as the number of audit hours spent on an engagement.
Nonetheless, a larger scale can also lead to a reduction in variable audit engagement costs. We rely on organizational learning theory from the management science literature to argue that industry scale leads to greater industry knowledge and subsequently to audit efficiency or lower audit hours, ceteris paribus. Organizational learning is a change in the organization’s knowledge that occurs as a function of cumulative experience and is often observed in labor intensive industries (Argote and Miron-Spektor 2011). This knowledge will result in a change in cognition, audit behavior, and/or performance and can manifest itself as a positive impact on efficiency (i.e., fewer audit hours) (Argote and Epple 1990; Boone et al. 2008; Cahan et al. 2011). The “learning curve” links
organizational learning and performance, and can be defined as the rate of improvement in the performance of a task (i.e., more efficiency) resulting from gains in cumulative production experience (Boone et al. 2008). Prior empirical studies provide ample evidence of organizational learning in both manufacturing (Argote and Epple 1990) and service firms, such as architectural-engineering or software-development firms (Darr et al. 1995; Pisano et al. 2001;
Reagans et al. 2005; Boh et al. 2007; Boone et al. 2008). Because audit firms are knowledge-intensive professional service firms and their production function relies on a substantial body of complex knowledge, we argue that
Accepted Article
learning curve effects also take place within these firms.5 Furthermore, a larger scale allows industry-specific knowledge to accumulate faster, enabling the audit office to move down its learning curve faster. This results in more efficient audits because fewer hours are required to perform audits of similar quality.
A further question in this context is at what level (individual partner, audit office, or audit firm level) is industry-specific knowledge shared within the audit firm. We follow recent studies that argue that a significant proportion of industry-specific knowledge is tacit and difficult to codify, which hinders the transfer of knowledge from the office level to the national level (Francis et al. 2005; Numan and Willekens 2012; Fung et al. 2012).
Furthermore, because experience, motivation, and know-how are all relevant in the audit profession, an IT-based system can never guarantee full knowledge transfers, and personnel interaction remains relevant (Vera-Munoz et al.
2006). Hence, we argue that industry learning curve effects will be strongest at the office level.6
To summarize, when applied to the audit context, organizational learning suggests that the cumulative expertise gained from servicing more audit clients in a specific industry (i.e., a higher industry office scale), results in enhanced organizational knowledge about the client’s industry and hence more efficient audits (i.e., fewer audit hours). Our first hypothesis is thus as follows:
HYPOTHESIS 1. Ceteris paribus, an auditor’s office industry scale is negatively associated with the number of audit hours spent on an engagement.
5 Because each audit engagement is unique and requires a high level of customization, learning does not occur through repetition of the same task, but through the generation of knowledge that can be implemented across all projects (Boone et al. 2008) or audit engagements. Note further that organizational learning in audit firms is stimulated by the profit sharing rules that exist in these firms. When partners share their individual knowledge with other partners, audit efficiency and effectiveness is enhanced (Chowet al. 2008; Vera-Munozet al. 2006). Hence, profit sharing rules that encourage knowledge sharing positively affect organizational learning (Chow et al. 2008; Liu and Simunic 2005; Vera-Munoz et al. 2006).
6There is also some evidence of national level industry-specific knowledge. For example, clients of industry specialists measured at the national level have lower discretionary accruals and higher earnings response coefficients (Balsam et al. 2003), pay higher fees (Craswell et al. 1995; Defond et al. 2000), and have a higher analyst ranking of disclosure quality in unregulated industries (Dunn and Mayhew 2004). The motivation for focusing on the office instead of the national level is threefold. First, it enables us to introduce industry fixed effects, which decreases the likelihood of the omitted variable problem at audit firm level. Second, we are able to define scale efficiencies in a more refined way because scale efficiencies can differ between offices in the same industry. Third, in line with recent studies about industry specialization, we expect that scale efficiencies are more difficult to transfer between offices (Francis et al. 2005).
Accepted Article
Economies of industry scale, efficiency discounts, and auditor market power (H2)
Next, we examine whether the audit firm’s market power affects the extent of efficiency pass-on to audit clients.
Prior studies document an association between client-auditor bargaining power and audit pricing (Fung et al. 2012;
Casterella et al. 2004; Mayhew and Wilkins 2003). Fung et al. (2012) and Casterella et al. (2004) take a client-side (i.e., demand-side) perspective by defining bargaining power as the importance of the client to the audit office, while Mayhew and Wilkins (2003) take an auditor-side (supply-side) perspective by arguing that industry leaders that sufficiently differentiate themselves from the second audit firm in the market (i.e., have at least 10 percent market share difference) have higher bargaining power. Although these studies provide important insights, the inferences are based on testing audit fees. Therefore, it is impossible to assess whether the magnitude of the observed fee discounts is due to passing on a larger part of the efficiency gain (thereby decreasing the auditor’s margin—in other words, resulting in a lower billing rate), or results from a decrease in audit hours (i.e., larger efficiency gains, thereby maintaining the auditor’s margin and thus the billing rate).
We argue that an auditor’s market power affects the pass-on of efficiencies. In markets with organizational learning, attracting new clients allows firms to move further down on their learning curve, thereby achieving greater efficiency and lower marginal production costs (Besanko et al. 2010; Cabral and Riordan 1994). Simultaneously, it also prevents competitors from moving down their own learning curves and achieving similar efficiencies. Hence, attracting new and retaining existing clients is of paramount importance in markets with learning effects because efficiency gains result in lower costs, allowing firms to set prices below those of their rivals. An audit firm’s client portfolio is thus an important asset and the firm with the largest portfolio has a strong competitive advantage. Given these market dynamics, firms can be expected to adopt aggressive pricing strategies, evident in, for example, completely passing on efficiency gains to clients via price discounts, in order to attract or retain these clients.
However, recent studies show that a full efficiency pass-on is not the only possible outcome in markets where learning curve effects occur (Besanko et al. 2010). The long run equilibrium outcome may result in a less than complete efficiency pass-on or even prices close to the monopoly level (Besanko et al. 2010) as the most efficient firms will be able to drive competitors out of the market, which could subsequently result in a less competitive market with fewer suppliers and higher prices (Dasgupta and Stiglitz 1988; Besanko et al. 2010). Consequently, the remaining firms in the market will gain greater market power and are eventually able to increase their prices above
Accepted Article
marginal cost (Dasgupta and Stiglitz 1988).7 The actual equilibrium outcome depends on the market power a firm has achieved (Dasgupt and Stiglitz 1988). The higher the audit firm’s market power, the more the client-auditor bargaining power shifts in favor of the auditor. Even if the client would switch to a competitor, the immediate efficiency gains realizable by this smaller market power competitor are likely to be too small to threaten the high market power firm in the short run. The high market power auditor will thus be less inclined to pass on all the realized efficiencies to clients. In this respect, Besanko et al. (2010, 475) conclude that “price competition between firms with similar stocks of know-how is intense but abates once firms become asymmetric.” The higher the firm’s market power, the less similar the audit firm is to its competitors. This results in more asymmetry and hence weaker price competition. This leads to our second hypothesis:
HYPOTHESIS 2. Ceteris paribus, the size of the audit fee discount attributable to economies of office industry scale (or the efficiency pass-on) is smaller when the auditor’s market power in the relevant industry market segment is higher.
3. Proprietary audit hours data from Belgium
As stressed earlier, we require data on audit engagement hours and fees to reliably test our predictions. In Belgium, between 1998 and 2000, audit firms were required to disclose the audit hours they spent on each audit engagement, together with the audit fee charged to the Belgian Institute of Auditors (IBR-IRE, Instituut van de Bedrijfsrevisoren – Institut des Reviseurs d’Entreprise). These data are not publicly available and were used by the Institute of Auditors to monitor low-balling and to safeguard audit quality. Since 2001, audit firms no longer have to disclose audit hours to the Institute of Auditors; that is, they only have to report the audit fees charged for each engagement.
Since 2007, audit fees became publicly available in clients’ annual reports. We were able to retrieve from the
7The degree to which driving competitors out of the market results in greater market power depends on entry barriers, among other factors. In addition, audit markets are generally not characterized as homogeneous markets. Indeed, spatial-differentiation theory applied to the audit market (Numan and Willekens 2012) suggests that firms can maintain prices above marginal cost (i. e., with partial efficiency pass-on) without losing market share if the firm is able to successfully differentiate itself from its
competitors, thereby increasing its market position relative to clients.
Accepted Article
Institute of Auditors the list of audit engagements of one Big 4 firm, including the audit hours per engagement as well as the audit fees charged for the year 2000.
As we use proprietary data from a Belgian Big 4 firm, it is relevant to describe some key features of the Belgian audit environment. In Belgium, all public and private firms exceeding certain size thresholds are required to appoint a statutory auditor (Vander Bauwhede and Willekens 2004).8 Because the hours and fee data of all audit engagements were disclosed to the IBR-IRE, we were able to compute scale measures that are based on all
engagements an audit firm has in an industry, and not only the audit engagements relating to public clients, as is the case in various prior studies that are based on Anglo-American datasets. There are significant differences, however, between public and private companies with respect to the regulatory environment, accounting practices, and required disclosure levels. Combining both public and private companies may therefore negatively affect our ability to make inferences. Because the proportion of publicly listed clients in a typical Belgian audit firm’s portfolio is very low, we removed the publicly listed clients from our sample.9
Note that private firm audits may differ substantially from public firm audits, and that this could create a bias against finding support for Hypothesis 1. First, from an audit production perspective, potential efficiency gains from industry scale are less likely if the required industry expertise to conduct audits in a private client setting is lower than in a public client setting. Second, privately held clients may have a lower demand for high-quality financial reporting (Ball and Shivakumar 2005; Cahan et al. 2011), which, in turn, may induce audit clients to value the audit less and therefore exercise higher audit fee pressure. These market characteristics create an environment where cost minimization strategies are likely to be more successful than audit quality differentiation strategies (Cahan et al. 2011). Such incentives to keep audit costs low may already lead to highly efficient audit practices making additional efficiency gains hard to achieve. Note further that audit fees were not publicly available in 2000, the year of our study. This means that client firms could not signal financial statement quality to stakeholders by
8Firms have to appoint an auditor if they average at least 100 employees (on yearly basis) or exceed two of the following criteria: (i) they employ 50 people, (ii) they have sales of 6.250 million EUR, and (iii) they have total assets of 3.125 million EUR.
9 We do this in order to obtain a more homogeneous sample of client firms. As a result, the reported effects of economies of scale and market power are not attributable to the differences in regulatory environments between public and private companies. We thank an anonymous reviewer for pointing this out. The results are qualitatively similar when including the seven publicly listed clients.
Accepted Article
paying higher audit fees. The absence of this signaling device allows for a clearer test of efficiency pass-on because it rules out client incentives for paying a higher audit fee. If signaling were present, client firms could have
incentives to resist efficiency pass-on because lower fees could be interpreted by the client’s stakeholders as reflecting lower financial statement quality.
From a supply perspective, the concentration of audit firms in the Belgian audit market is typically lower than in U.S. markets (Dutillieux 2009; Willekens and Achmadi 2003; Schaen and Majoor 1997; Weets and Jegers 1997). Dutillieux (2009) reports that the mean number of audit firms active in a 2-digit SIC industry is 42 in Belgium compared with 18 in the United States. Hence, in contrast to the U.S. audit market, the Belgian market has more variation in audit supplier concentration and auditor market power across industries. Note, however, that this could also be partially due to the fact that a more complete list of audit engagements (both public and private) is included in the Belgian databases.
Finally, it is relevant to note that Belgium is geographically less dispersed than the United States and Australia, countries on which many other auditor industry expertise studies focus. In 2000, the average Belgian Big 5 audit firm had about four to five offices, not all big accounting firms had offices in every major city, and audit firms were competing for the same clients from different offices. A single office can even have clients from across the country. This feature hinders the definition of market segments and measuring market power at the city-specific level.
4. Research design
Measures of auditor industry scale
We made a number of motivated choices when specifying our measures of auditor industry scale. We follow Fung et al. (2012) and construct a first measure of office industry scale, defined as SCALE, based on the number of clients.
We start by calculating the number of clients in each office-industry pair. Thereafter, the number of clients is ranked across all office-industries. We then measure SCALE as the percentile rank of the number of clients where higher values of SCALE reflect more clients in the office-industry. Hence, SCALE ranges from 0.01 to 1. Next, we specify a second measure of industry scale similar to studies using portfolio share measures of industry specialization
(Krishnan 2001; Krishnan 2003; Neal and Riley 2004). Portfolio share based measures fit the underlying theory used to motivate our hypotheses because they capture the resources invested in each industry within an audit office. More
Accepted Article
specifically, we define a within-office industry portfolio share measure based on the square root of total assets audited in each client industry.10 We construct a dummy variable, defined as TOP3, which equals one for the three largest portfolio share industries in each office, and zero otherwise (Neal and Riley 2004). Additionally, to receive a value of one for TOP3, we require that the office has at least five clients in each office-industry because the
operating scale is otherwise unlikely to be large enough to result in organizational learning and thus scale economies. We use 2-digit SIC codes to define client industries consistent with prior literature (e.g., Francis et al.
2005).
We link each client to the office from which it is audited in the following manner. First, we identify the partner auditing each engagement. Second, we match each partner to an audit office by using their registered professional location as it appears in the Instituut van de Bedrijfsrevisoren (IBR) membership list, as published in Het Belgisch Staatsblad in 2000. Whereas prior audit studies assume one office per geographical location (Francis et al. 2005; Numan and Willekens 2012), this procedure allows us to exactly match all clients with the corresponding audit office, irrespective of the clients’ location.
Measure of auditor market power and relevant market segments in Belgium
The literature has not reached a consensus about which measures are better for capturing audit firms’ market power.
We use the absolute value of market share distance to the closest competitor, defined as DISTANCE, as the basis for our measure of market power (Numan and Willekens 2012).11 This measure is derived from spatial competition research in industrial organizations. We believe that this is a valid market power measure in this setting for the following reasons. First, we need a firm-specific measure of industry market power because our sample consists of
10We use client assets rather than client fees because any measure based on audit fees is, by construction, correlated with the dependent variable in the fee model and leads to spurious correlations. The formula of the portfolio share of each office-industry is described below where i represents the audit client, j the industry in which the client operates, and o the audit office that audits the client. In addition, n reflects the number of clients of the audit office in the industry of the client, while m reflects the number of clients of the audit office across all industrie𝑠: (∑𝑛𝑖=1√𝐴𝑆𝑆𝐸𝑇𝑆𝑖𝑗𝑜) ∑𝑚 √𝐴𝑆𝑆𝐸𝑇𝑆𝑖𝑜
⁄ 𝑖=1 .
11 We measure DISTANCE in the following way: |min (𝑀𝑆𝑖− 𝑀𝑆𝑖+1; 𝑀𝑆𝑖− 𝑀𝑆𝑖−1)| where MS represents market share, i represents the sample firm; i+1 is the closest firm that has a market share that is higher than the sample firm, and i−1 is the closest firm that has a market share that is lower than the sample firm:
Accepted Article
(all) audit engagements of one audit firm. DISTANCE captures the extent of differentiation between an audit firm and its closest competitor and thus varies between firms within the same industry. This measure is therefore better suited to the purposes of this study than is the Herfindahl-Hirschman index, which assumes that all firms within an industry have the same market power. Second, following Besanko et al.(2010) we need a measure of competition that captures how (a)symmetric competing audit firms are. The market share distance measure reflects how (a)symmetric or (dis)similar the audit firm is with its closest competitor and is thus consistent with the model used by Besanko at al. (2010).
As the distribution of DISTANCE is highly skewed (see also Table 3), we construct a dummy variable HIGHMP to differentiate between industry segments in which an audit firm has high and low market power.
HIGHMP is equal to one when DISTANCE is higher than its p75-value (0.038), and zero otherwise. By constructing a dummy variable, we are also able to split our sample based on the audit firm’s market power in order to test Hypothesis 2.
To calculate an auditor’s market share we need to identify the relevant audit market segments in which auditors compete. We follow recent studies that delineate audit markets based on industries within a geographical area and construct these measures based on client location (Francis et al. 2005; Reichelt and Wang 2010; Numan and Willekens 2012). In the Belgian context, there are three regions (the Flemish Region, the Walloon Region, and the Brussels Capital Region) that are relevant geographical areas within which auditors compete. This geographical classification into three regions is motivated by the following arguments. First, both the OECD and Eurostat construct their regional economic indicators based on these regions.12 Second, besides cultural and language differences between these regions, important parts of economic policy fall under the purview of the regional governments in Belgium. Hence, significant economic and competitive differences may exist between the regions.13 Hence, the relevant market segments are the different industries within each of these three regions.
12 Eurostat, for instance, uses the Nomenclature of Territorial Units for Statistics (NUTS) to provide a single, uniform breakdown of territorial units. The three Belgian regions correspond to NUTS1, which denotes major socio-economic regions (European Commission 2015).
13Note that dividing each region into smaller areas (like cities or MSAs) to construct office- or city-specific market segments is not appropriate for the following reasons. First, the clients audited by a single audit office are not all located within the same city or the same province or geographical location. Some clients are audited by an office that is further away than the closest office of
Accepted Article
Control variables
In line with previous research (Hay et al. 2006), we include several other controls in the audit-hours, audit-fee and billing rate models. First, consistent with Bae et al. (2016), we construct a market share-based measure of industry specialization. We include a dummy variable equal to one when the audit firm is an industry leader in the region with a market share distance of at least 10 percent, and zero otherwise (DOM_LEADER). Second, we control for several client-side characteristics such as size (LN_ASSETS and LN_SALES), complexity (SUBSIDIARIES,
PUBLIC), audit risk (RECINV), financial risk (QUICK, DEBT, LOSS, ROA), and the provision of additional services (LN_NAS). We also include the following engagement-specific characteristics: the initial mandate (i.e., lower or equal to a tenure of 3 years) (INITIAL), the timing of the audit (BUSY) and the presence of a joint audit (JOINT).
Since our sample firms had taken over some smaller audit firms and offices, we control for whether the client was attracted due to the acquisition of a rival firm given that these clients may differ from the other clients (ACQUIR).
Finally, we also include office and industry fixed effects to control for any audit-office and industry differences other than office industry scale in audit hours and audit fees (Stein et al. 1994).14,15
Tests of the hypotheses
To test our Hypothesis 1, we specify an audit engagement hours model which includes the following explanatory variables: (i) a variable capturing auditor industry scale; (ii) a variable capturing the auditor’s market power; and (iii) a number of control variables similar to those used in prior studies (Hay et al. 2006):
LN_HOURS = α0 + α1SCALE/TOP3 + α2HIGHMP + α3DOM_LEADER + α4LN_ASSETS + α5LN_SALES + α6RECINV + α7QUICK + α8DEBT + α9ROA + α10BUSY + α11LOSS + α12INITIAL + α13JOINT +α14ACQUIR + α15LN_NAS + α16SUBSIDIARIES +Office and industry fixed effects+ ε . (1)
the same firm, while in other locations clients tend to divide themselves evenly across two different offices of the same a udit firm. It is unlikely, given their shared brand name, that these offices would compete against each other. Second, not all big accounting firms are located in the same city and most cities do not have an office of each big accounting firm. Third, the geographical distance between cities within Belgium is short; one can cross the entire country in three hours by car. The distance between the two most distinct audit offices in our sample is about 200 km. These characteristics make establishing geographical boundaries for office-specific markets challenging and any choice would necessarily be ad hoc.
14 For instance, we find that audit fees of the Brussels office are higher than those of other offices.
15 If firm-level expertise leads to more efficient audits, then industry-specific fixed effects should also capture these effects.
Accepted Article
Definitions of the variables in equation (1) are provided in Table 1.
We test Hypothesis 1 using three specifications of audit engagement hours: (i) total hours (LN_HOURS), (ii) partner hours (LN_PARTHOURS) and (iii) staff hours (LN_STAHOURS). The first hypothesis predicts a negative α1-coefficient for the industry scale variables (SCALE and TOP3). To examine whether efficiency gains are passed on to clients, we also estimate equation (1) with audit fees as dependent variable. As the error terms of these different models are likely correlated, we estimate these models using Seemingly Unrelated Regressions (SUR), a methodology that is also used in other recent accounting studies (Fresard 2010; Byard et al. 2011; Armstrong et al.
2012).16 Note that we also test whether a larger scale allows the firm to divide fixed costs over more clients by estimating a billing rate model, where the dependent variable is defined as the natural logarithm of the total audit fee divided by the total hours (BILLRATE), and the same set of explanatory variables as in equation (1). When the coefficient of SCALE and TOP3 is significantly negative in the billing rate model, this implies that fixed costs are spread out over more clients, leading to lower average costs and a lower billing rate.
To test the second hypothesis, we specify the following regression models:
LN_HOURS and LN_FEE/BILLRATE = β0 + β1SCALE/TOP3_LOWMP+ β2SCALE/TOP3_HIGHMP + β3HIGHMP + β4DOM_LEADER + β5LN_ASSETS + β6LN_SALES + β7RECINV + β8QUICK + β9DEBT + β10ROA + β11BUSY + β12LOSS + β13INITIAL + β14JOINT + β15ACQUIR + β16LN_NAS
+β17SUBSIDIARIES + Office and industry fixed effects+ δ . (2) We split SCALE and TOP3 into two distinct variables depending on the value of HIGHMP. More
specifically, the variable SCALE/TOP3_LOWMP equals the value of SCALE or TOP3 if HIGHMP equals zero, and zero otherwise. In contrast, the variable SCALE/TOP3_HIGHMP equals the value of SCALE or TOP3 if HIGHMP equals one, and zero otherwise.
We test Hypothesis 2 in two ways. First, we test equation (2) by estimating two distinct dependent variables, audit fees (LN_FEE) and audit hours (LN_HOURS), simultaneously using SUR. This allows us to
16 SUR estimates the different models simultaneously and accounts for the correlations of error terms across equations, in contrast to OLS regression. Given that audit hours and audit fees are not independent, their error terms are likely to be correlated.
Analysis reveals significant correlation of the error terms across the audit hours and audit fee model (e.g., r = 0.66, p-value <
0.01), and the Breusch-Pagan test rejects the null hypothesis that the audit hours and audit fee equations are independent (χ² = 393.03, p-value < 0.01) based on the base model without SCALE, TOP3, or MPOWER.
Accepted Article
compare the coefficients of the fee model with the coefficients of the audit hours model using a Chi-square test of equal regression coefficients (Fresard 2010; Byard et al. 2011; Armstrong et al. 2012).17 The second hypothesis is supported when the Chi-square test of equal regression coefficients indicates that the coefficient of
SCALE/TOP3_HIGHMP has a significantly lower value in the audit hours model than in the audit fee model, indicating that the percentage decrease in audit fees is significantly smaller than the corresponding percentage decrease in audit hours. Second, we use billing rates (see Bae et al. 2016; Redmayne et al. 2010) as the dependent variable (BILLRATE) and estimate equation (2). Hypothesis 2 is supported when SCALE/TOP3_HIGHMP is significantly positively associated with BILLRATE.
5. Sample selection
We use a proprietary dataset of 1,258 individual engagements from one Belgian Big 4 firm for the year 2000 containing audit hours as well as audit fees. We merge these audit hours data with the clients’ financial statement data found in the database BEL-FIRST.18 From the initial sample, we remove 228 observations that were missing data for audit hours (62), control variables (122), or audit fees (44). We also exclude 120 observations from the financial and real-estate industries (SIC code 6000–6999) and 7 public clients. Finally, in order to calculate market share measures that are meaningful, we exclude industries with less than five clients at the regional level, as the market share measures in those industries (and hence HIGHMP and DOM_LEADER) can be subject to considerable measurement error. This leads to a loss of 8 observations, resulting in a final sample of 895 observations. This sample size is significantly larger than those used in prior audit hours studies (O’Keefe et al. 1994; Schelleman and Knechel 2010).
We also use BEL-FIRST to gather information about the total audit market in Belgium, using all clients that were required to appoint a statutory auditor in 2000. Our initial sample contains 15,733 observations. After
17In contrast to OLS regression, SUR estimates the audit hours and audit fee models simultaneously, accounts for the correlations of error terms across equations and allows for a Chi-square test of equal regression coefficients across equations. Comparing coefficients across a fee and an hours regression model, in which both dependent variables are measured as a logarithmic transformation, is equivalent to testing the null hypothesis that the percentage change in hours attributable to office industry scale is equal to the percentage change in audit fees attributable to office industry scale (a formal proof of this statement is available upon request).
18 BEL-FIRST is a product of Bureau Van Dijk Electronic Publishing (www.bvdep.com). It contains financial-statement information from all the firms in Belgium and Luxembourg that are required to file financial statements.
Accepted Article
removing observations that had missing values for total assets (299 observations), for the appointed audit firm (2,354 observations), or for the SIC code (1,807 observations), we arrive at a final sample of 11,343 observations, which we use to calculate the market power proxy. An overview of our sample-selection procedure can be found in Table 2.
[Insert TABLE 2 about here]
6. Results
Descriptive statistics
Table 3 provides the descriptive statistics for the variables we use in our empirical tests. Panel A shows that the average (median) client pays audit fees of 15,006 EUR (8,393 EUR) and that the average (median) audit
engagement is performed using approximately 163 (98) audit hours, 10 (4) partner hours and 153 (92) staff hours.
Panel B shows that the average (median) value of SCALE is 0.780 (0.880). Our second office industry scale measure, TOP3, equals one in 38.7 percent of the observations. The descriptive statistics for the control variables in panel C show that the average (median) value of LN_ASSETS is 19.376 (19.262), the average debt-to-assets ratio (DEBT) is 74.9 percent, and the mean (median) return on assets (ROA) is 0.020 (0.027), which is similar to prior research (Schelleman and Knechel 2010). Finally, panel D of Table 3 shows substantial variation in the distance to the closest competitor. The average (median) distance is 3.4 percent (1.3 percent), with a maximum distance of 85.7 percent. Overall, descriptive statistics for this variable are equivalent to those reported in Dutillieux (2009), who also uses the Belgian setting.
[Insert TABLE 3 about here]
In Table 4, we provide Pearson and Spearman correlations. Our audit office industry scale variables are negatively correlated with LN_HOURS (SCALE: –0.0651; TOP3: –0.0477), LN_PARTHOURS (SCALE: –0.0712;
TOP3: –0.0764), LN_STAHOURS (SCALE: –0.0624; TOP3: –0.0481) and LN_FEE (SCALE: –0.0772, TOP3: –0.0625), albeit mostly insignificantly at the 5 percent level. In addition, the audit office industry scale
Accepted Article
variables are not associated with the firm’s market power (SCALE: –0.0095; TOP3: 0.0212).19 Finally, correlations between the other variables are acceptable and in line with previous research, the highest absolute correlation of 0.7063 being between LN_SALES and LN_ASSETS.
[Insert TABLE 4 about here]
Regression results: Test of Hypothesis 1
Table 5 reports the results of estimating equation (1) with adjusted standard errors for heteroscedasticity and clustered standard errors by industry and tests Hypothesis 1.20 We find a significant negative coefficient for both office industry scale measures in the audit hours regression (SCALE: –0.558, p-value < 0.05; TOP3: –0.155, p-value
< 0.10), in support of Hypothesis 1. In economic terms, we find that an increase of one standard deviation in SCALE results in an efficiency gain of 13.45 percent. The engagements of clients of one of the largest three portfolio industries of an audit office are performed with 14.36 percent fewer audit hours. In addition, we find that staff hours (SCALE: –0.581, p-value < 0.05; TOP3: –0.155, p-value < 0.10) and partner hours decrease with industry scale, albeit the latter only using the TOP3 proxy (coeff: –0.252, p-value < 0.05). These results support the first hypothesis that office industry scale results in more efficient audits, and show that higher industry scale reduces the variable costs of the engagement.
Note that the results show that market share-based industry expertise is associated with higher audit hours (DOM_LEADER). We thus find a negative effect of office industry scale on audit hours, in addition to a positive effect of market share-based industry specialization (Bae et al. 2016). This confirms that market share-based and portfolio share-based measures (which are by construction correlated with scale) of industry specialization may
19Theoretically, a positive mechanical correlation between our measures of SCALE and DOM_LEADER is plausible. Indeed, all else equal and if all industries would be of equal size, being the leader in an industry requires a higher market share than competitors, which would increase the operating scale in that industry. However, since the reported correlations between SCALE and DOM_LEADER are negative we do not believe that such mechanical correlation drives our results.
20 Depending on the proxy of office industry scale, we find an adjusted R² of about 31 to 42 percent for the audit hours model, which is lower than reported in previous audit hours studies in U.S. and Australian settings (Bellet al. 2008; Schelleman and Knechel 2010) but similar to audit-fee models that use Belgian data (Dutillieux et al. 2009).
Accepted Article
capture two different aspects of industry knowledge and may have a differential effect on audit hours and/or audit fees. This finding adds to the debate on the measurement of industry specialization (Minutti-Meza 2013; Audousset- Coulier et al. 2015). For audit offices that both have a high industry scale and enjoy a high market share, opposite effects occur simultaneously, which could explain the seemingly contradictory results found in the literature on industry specialization effects. The results for the control variables are generally consistent with those of previous studies.
[Insert TABLE 5 about here]
Table 6 investigates the effect of office industry scale on audit fees and billing rates. The results show that audit fees are negatively associated with office industry scale (SCALE: –0.408, p-value < 0.05), although the coefficient of TOP3 in the fee model is insignificant (TOP3: –0.164, p-value > 0.10). Furthermore, the billing rates are unaffected by office industry scale (SCALE: –0.205, p-value > 0.10; TOP3: –0.008, p-value > 0.10). We thus find that the audit fee discount associated with scale, as also reported in Fung et al. (2012), is not driven by the division of fixed costs over more clients (which would result in lower average costs and lower billing rates). 21,22 This finding further also implies that efficiency gains are fully passed on to clients. Note that this conclusion is also supported by the insignificant test of equal regression coefficients (unreported) between the audit hours and the audit fee model (SCALE: p-value = 0.319; TOP3: p-value = 0.892).
[Insert TABLE 6 about here]
21 Note that, in the absence of data on these fixed costs, we cannot rule out the possibility that fixed costs are spread out over more clients but that these cost savings are not passed-on to clients.
22 With respect to DOM_LEADER, we find that audit fees are higher (coeff: 0.387, p-value < 0.05) in line with the findings reported in Bae et al. (2016), who rely on South Korean data. We, however, do not find higher billing rates associated with DOM_LEADER (p-value > 0.10), while Bae et al. (2016) find a lower billing rate for dominant leaders. This difference is likely caused by the different types of firms (private versus public) and differences in the institutional environments between Belgium and South Korea.
Accepted Article
Regression results: Test of Hypothesis 2
In Table 7, we report in panels A and B the results of our test of Hypothesis 2 using SCALE and TOP3 as our measure of office industry scale, respectively. Our main variable of interest is SCALE/TOP3_HIGHMP, which captures the effect of office industry scale when auditor market power is high. We document a higher billing rate associated with office industry scale when market power is high (SCALE_HIGHMP: 0.380, p-value < 0.05 and TOP3_HIGHMP: 0.220, p-value < 0.01). This suggests that the audit fee discount (and hence efficiency pass-on) is smaller when the audit firm has market power, in line with Hypothesis 2. This finding is further supported by the significant Chi-square test of equal regression coefficients for both TOP3_HIGHMP and SCALE_HIGHMP (p-value
< 0.05), which shows that the realized efficiency gain (SCALE_HIGHMP: –0.772, p-value < 0.01 and
TOP3_HIGHMP: –0.531, p-value < 0.01) is not fully reflected in the decrease in audit fees (SCALE_HIGHMP:
–0.410, p-value < 0.10 and TOP3_HIGHMP: –0.298, p-value < 0.01). Hence efficiency gains are only partially passed on when the audit firm’s market power is high.
In contrast, we find no evidence that the billing rate is higher when market power is low (SCALE_LOWMP:
0.132, p-value > 0.10 and TOP3_LOWMP: –0.078, p-value > 0.10), which is also supported by our test of equal regression coefficients (p-value > 0.10). This finding shows that efficiency gains, when realized, are fully passed on to clients when the audit firm has low market power. Taken together, these results confirm our second hypothesis, as the extent of pass-on decreases from a full pass-on to a partial pass-on when the audit firm’s market power is high.
An untabulated Wald test reveals that the coefficient for TOP3_HIGHMP is significantly different from TOP3_LOWMP in the audit hours model (p-value < 0.01). This suggests that efficiency gains are larger when the audit office has a sufficiently large scale and market power is high. While we did not ex ante hypothesize such an effect, this could indicate that audit firms seeking to differentiate themselves from their closest competitors focus more on their industry knowledge and organizational learning processes, allowing them to benefit more from economies of scale.23 The Wald test, however, shows no significant difference between SCALE_LOWMP and SCALE_HIGHMP in the audit hours model (p-value > 0.10). Hence, we are hesitant to draw strong inferences.
23This result is also consistent with Fung et al. (2012), who report a similar finding: a significantly lower interaction effect of economies of scale and industry leadership on audit fees. Because they only analyzed audit fees, they interpreted this effect as
Accepted Article
The results for the other control variables are generally in line with expectations. The billing rate is lower for larger (p-value < 0.10), profitable clients (p-value < 0.05), and those recently acquired through audit firm merger activity (p-value < 0.10). Interestingly, the results show the additional audit effort (p-value <0.05) associated with larger inventories or receivables is not reflected in the audit fee (p-value > 0.10), resulting in a lower billing rate (p- value < 0.05). Similarly, there is evidence of a fee premium during the busy season, although the effect is
insignificant in the billing rate model using TOP3.
[Insert TABLE 7 about here]
Supplementary analysis: The association between industry scale and audit quality
Our results suggest that office industry scale leads to efficiency gains. However, a decrease in the time spent on engagements could be risky because performing fewer audit hours could imply that the minimum audit quality is no longer supplied. To test whether our office industry scale measures are associated with differences in audit quality, we estimate a model in which we explain absolute, positive and negative abnormal accruals as a function of office industry scale and of a number of control variables typically used in prior international (Reichelt and Wang 2010) and Belgian earnings-management studies (Dutillieux 2009).24 We estimate performance-adjusted abnormal discretionary accruals using the population of all Belgian firms for each industry with at least 20 observations (i.e., out of sample) (Reichelt and Wang 2010; Simnett et al. 2016).25 Due to a lack of data for estimating the level of
industry specialists passing more of the efficiencies on to their client. The implicit assumption in their study is that the magnitude of the efficiency gain does not differ between industry specialists and non-specialists. Yet, our results with respect to TOP3 suggest that efficiency gains themselves may actually increase, while the pass-on decreases.
24 We also estimated abnormal working capital accruals. The untabulated results are qualitatively similar.
25 Abnormal accruals are calculated as the portion of total accruals (calculated as net income from continuing operations minus operating cash flow) that are “unexpected,” given the firm’s characteristics. Hence, abnormal accruals are a proxy for the extent to which management uses accruals to adjust earnings upward (or downward). We initially start with a sample of the 15,733 firms required to appoint a statutory auditor in 2000. We delete observations with missing values for LN_ASSETS (– 229), industry codes (– 1,807), insufficient data for calculating total accruals (– 914), and insufficient prior-year data for constructing the control variables (– 1,627). After truncating scaled total accruals at the top and bottom 1 percent, the final sample consists of 10,934 firms.
Accepted Article
earnings management, only 825 of the initial sample of 910 could be used.26 Panel A in Table 8 displays the results using SCALE as the measure of office industry scale, while panel B displays the result of TOP3. The adjusted R2 is about 7.2 percent, which is comparable to prior Belgian abnormal accrual studies (Dutillieux 2009). With respect to the variables of interest, the coefficients of our measures of office industry scale (SCALE: coeff: 0.0024, p-value >
0.10 and TOP3, coeff: 0.002, p-value > 0.10) are not significantly associated with the absolute, negative or positive discretionary accruals. Hence, efficiency gains due to office industry scale are not associated with a reduction in audit quality.
[ Insert TABLE 8 about here]
7. Sensitivity analyses
Alternative measures of office industry scale
To test the robustness of our results, we use a scale measure by including a dummy equal to one for the four largest portfolio industries. Overall, the untabulated results are qualitatively similar to the main models.
Alternative measures of audit firm market power
We construct dummy variables based on different cutoffs of DISTANCE. Specifically, we develop different indicator variables based on a market-share distance to the closest competitor of 1 percent, 2 percent, 3 percent, and 4 percent. In line with the main analysis, we split SCALE and TOP3 based on these dummies. The billing rate models shows a higher billing rate when using cutoffs of 2 percent, 3 percent and 4 percent, while this is not the case for a cutoff of 1 percent. In sum, these results suggest that the audit firm needs at least a 2 percent market share distance to enjoy market power vis-à-vis the client.
Client size
To test whether the results are driven by large or small private clients, we split the sample based on the median value of clients’ total assets. The untabulated results show that the results are mainly driven by the large private clients as
26 We have 873 observations with sufficient data to estimate absolute discretionary accruals, but the additional requirement of at least 20 observations per industry and the additional controls needed for the discretionary accrual models decreases the sample to 825.
Accepted Article
the results for this subsample are similar to the main results. In contrast, we do not find efficiency gains associated with office industry scale for the small clients.
Client bargaining power
Testing the impact of client bargaining power (Casterella et al. 2004), we run a split sample analysis based on the median value of client sales in each office-industry.27 The results suggest that client bargaining power dominates audit firm market power as the billing rate is only higher when the audit firm has market power and the client has low bargaining power. When the client’s bargaining power is high, all realized efficiencies are passed on regardless of the market power of the auditor. This is evident in the insignificance of the test of equal coefficients for the variables SCALE_HIGHMP and TOP3_HIGHMP in the subsample of clients with high bargaining power. In the subsample of clients with low bargaining power, the opposite situation holds, as the test of equal coefficients for those two variables is significantly different from zero, which suggests a partial pass-on.
Sample
To confront the concern that client characteristics could be correlated with SCALE/TOP3 we create a matched sample using a propensity score retrieved on all control variables. Results of the main analyses are confirmed. We find a significant negative coefficient for the TOP3 variable in the audit hour regression, while we do not find a higher billing rate. We do however find higher billing rates when the audit firm’s market power is high.
8. Conclusions and limitations
In this article, we examine whether office industry scale is associated with more efficient audits and whether the extent of pass-on of those efficiencies from the auditor to its client depends on the audit firm’s market power. Our results indicate that office industry scale is negatively associated with audit hours, but does not significantly affect
27 We could not use the proxy for bargaining power developed by Casterella et al. (2004), as measured by the natural log of total sales for client i divided by the natural logarithm of total sales for all clients audited by the auditor in the same 2-digit industry and office. The reason is that is that inspecting the split sample descriptives based on the median of POWER as in Casterella et al.
(2004) reveals that TOP3 accounts for 71.81 percent of observations in the low client bargaining power sample, but only 5.5 percent of observations in the high client bargaining power sample. This would significantly affect the validity of the results found.
